Gary, > > The reason I asked the question: > > "Could you share with us what you meant by the statement: "a V.90 modem > with marginal return loss may not work very well. " > > was not from the point of return loss or balance that's a simple problem to > solve. The question was asked to understand how it effects V.90 > performance. As I stated in a previous email most analog modem chips will > function at maximum performance with a return loss of 6 - 7 dB. V.90 will > not.
In "most" designs the network matching is derived from the hybrid forwards through the line transformer and not from how well the modem chipset implements the modulation scheme. > > I'm humble enough to know that I will be learning all my life. As a > consequence I am great believer in asking questions. That's how I learn > many things. As am I. If this correspondence brings other input to this matter (such as telecom wiring via fence wires) I too will find this a learning experience. > > In my opinion simply asking for a second opinion does not exhibit > ignorance. It exhibits a mind open to different fields of view and > interpretations. Even though I bill myself as a telecom consultant there > is no claim to know all. I doubt anyone does. Then perhaps one should be a little more careful in how questions are presented as your point concerning modem performance sounded to me as a statement. Concerning an all encompassing knowledge, I heartily agree that no one person could know everything needed - hence the reason a forum such as this exists! > > The transmission theory classes I struggled through did not cover designing > equipment for subscriber loops used as a digital transmission medium. The > problems, to name a few, are standing waves (return loss), holding > circuits, line current, a mix of wire gauges, loading coils, capacitive > reactance, variable loop lengths from 200 to 25K meters and in some cases > barb wire. No matter how well it's modeled subscriber loops are very > hostile medium for V.90 and xDSL. I agree fully with this and would use this argument to state the need for modems of any description to match the reference network as closely as possible, in order that vaguaries of the actual installation have only minor detrimental effect on performance. > > In the last 30 years most of us have solved the subscriber loop problems > effecting analog modems. It's fairly easy to get return loss figures > across the band in the 25 to 40 dB range at 600 ohms and 18 to 25 dB with a > complex impedance. But, in my opinion V.90 is a special case and xDSL even > more so. > > Anyone that has designed transmission equipment that pushes bandwidths to > limit and beyond understands it is fairly straight foward to design for > perform at maximum throughput from point A to point B. But, when the > equipment has to perform at maximum throughput on millions of subscriber > loops that's a much tougher problem. I agree that devices such as V.90 modems utilise as much of the available bandwidth as possible to achieve maximum performance in the presence of sometimes indifferent (for what ever reason) network conditions. Therefore, I consider that maximising transmission efficiency by spending a little more time during the development stage pays dividends later in terms of reduced testing fees and fewer calls to support lines. xDSL modems bring their own set of luggage with them. The (much) greater amount of power needed to ensure transmission increases crosstalk in the exchange equipment and using a POTS splitter adds to the problems too! Throw in the need to use *all* the available bandwidth available above approximately 8kHz and network matching becomes ever more important. > > I hope that wasn't too defensive or pedantic. Not at all, its enjoyable to discuss these matters with someone who shares an interest! Gary
